ParseBench — hosted parsers
The published ParseBench leaderboard: twelve hosted parsing pipelines scored on tables, charts, content faithfulness, semantic formatting, and visual grounding.
12 systems · ~2,000 human-verified enterprise pages · last run August 6, 2026 · ParseBench (arXiv 2604.08538) · What ParseBench measures.
Winning result
Quality — 84.88 (LlamaParse Agentic)
Value — 0.28¢ (Databricks AI Parse)
Speed — not scored on this board.
Leaderboard
| # | System | Overall | Tables | Charts | Faithfulness | Formatting | Grounding | Cost / page |
|---|
| 1 | LlamaParse Agentic | 84.88 | 90.74 | 78.11 | 89.68 | 85.24 | 80.62 | 1.25¢ |
|---|
| 2 | Reducto (Agentic) | 72.97 | 80.42 | 73.40 | 86.37 | 57.60 | 67.07 | 4.76¢ |
|---|
| 3 | LlamaParse Cost Effective | 71.89 | 73.16 | 66.66 | 88.02 | 73.04 | 58.56 | 0.38¢ |
|---|
| 4 | Reducto | 67.83 | 70.33 | 56.99 | 86.37 | 56.75 | 68.71 | 2.38¢ |
|---|
| 5 | Extend (Beta) | 67.83 | 85.93 | 40.42 | 85.03 | 59.49 | 68.28 | 2.50¢ |
|---|
| 6 | Azure Document Intelligence (Layout) | 59.64 | 86.00 | 1.56 | 84.93 | 51.93 | 73.78 | 1.00¢ |
|---|
| 7 | Extend | 55.75 | 85.05 | 1.59 | 84.08 | 47.36 | 60.67 | 2.50¢ |
|---|
| 8 | Databricks AI Parse | 52.22 | 83.67 | 0.00 | 88.25 | 55.25 | 33.91 | 0.28¢ |
|---|
| 9 | Google Cloud Document AI | 50.39 | 55.10 | 1.44 | 83.65 | 50.51 | 61.26 | 1.00¢ |
|---|
| 10 | AWS Textract | 47.88 | 84.58 | 5.97 | 74.76 | 3.71 | 70.36 | 1.50¢ |
|---|
| 11 | LandingAI | 45.23 | 73.72 | 10.88 | 88.60 | 27.87 | 25.08 | 3.00¢ |
|---|
| 12 | Firecrawl | 31.08 | 55.88 | 0.00 | 74.37 | 25.16 | 0.00 | 0.90¢ |
|---|
- okraPDF is not scored on this board — it covers the hosted pipelines the published leaderboard runs.
- Cost per page is list price at the time of the run, not a measured spend.